1. Field of the Invention
The present invention relates to an adaptive equalizer, and more particularly to an adaptive equalizer which can compensate a signal distortion on a transmission path and which uses a delayed decision feedback sequence estimator.
2. Description of the Related Art
Generally, in a digital mobile communication, interference occurs between signal codes on a transmission path due to multi-path. The interference is a factor of transmission performance degradation. As one effectual measure to compensate the interference between the signal codes, equalizers of various types are used such as a maximum likelihood sequence estimation (MLSE) type.
In the digital mobile communication, the transmission path characteristic changes every moment in accordance with the movement of a mobile station. It is an adaptive equalizer that updates an impulse response of the transmission path in accordance with the transmission path characteristic change. At this time, various algorithms such as a least mean square (LMS) algorithm and a recursive least square (RLS) are used as the adaptive algorithm. It should be noted that the maximum likelihood sequence estimation is described by Horikoshi in xe2x80x9cWaveform Equalization Technique for Digital Mobile Communicationxe2x80x9d (pp. 77-92, published from TRICEPS). Also, the adaptive algorithm such as the LMS algorithm is described on pages 33-45 in the same publication.
Conventionally, an adaptive equalizer using the adaptive maximum likelihood sequence estimation is disclosed in Japanese Laid Open Patent Applications (JP-A-Heisei 5-152893 and JP-A-Heisei 5-152894). FIG. 1 shows a system employing a conventional adaptive equalizer. Referring to FIG. 1, after a reception signal is temporarily stored in a memory 60, the reception signal is supplied to a Viterbi algorithm processing section 70 through a matched filter 61. The Viterbi algorithm processing section 70 carries out the maximum likelihood sequence estimation to the reception signal and outputs an estimation result. It should be noted that the reception signal has a frame format shown in FIG. 3 and is composed of a known training signal and an unknown data signal.
An impulse response of a transmission path is used in the maximum likelihood sequence estimation by a transmission path estimating section 80. When the reception signal is supplied from the memory 60, the transmission path estimating section 80 first determines initial values by use of an adaptive algorithm such that a difference is made small between the reception signal and a replica signal. The replica signal is obtained by convoluting at the receiving end a training signal having a predetermined known value and an impulse response during the reception of the training signal. Also, the transmission path estimating section 80 updates the initial values during the reception of the data signal by use of the adaptive algorithm. The adaptive algorithm functions such that a difference is made small between the reception signal and the replica signal obtained by convoluting the impulse response and the estimation result from the Viterbi algorithm processing section 70.
The training signal or the estimation result from the Viterbi algorithm processing section 70 are multiplied with the impulse response components h0n to hLn by multipliers 82-0 to 82-L through delay elements 81-0 to 81-(Lxe2x88x921). The outputs of the multipliers 82-0 to 82-L are added by an adder 83. The output of the adder 83 is the replica signal to imitate the reception signal. An error signal is determined by an adder 84 to indicate a difference between the replica signal and the reception signal. A calculating section 85 calculates impulse response estimation values {Ehj} (j=0, . . . , L) using the error signal based on the adaptive algorithm and outputs them to the matched filter 61 and the Viterbi algorithm processing section 70.
In the above conventional example, the calculating section 85 uses the least mean square algorithm having the function to estimate a plurality of impulse response components using a plurality of parameter correction coefficients. A calculating section 71 has the function to determine optimal correction coefficients based on pathmetric result obtained as the result of the maximum likelihood sequence estimation using the plurality of impulse response components.
By the way, the maximum likelihood sequence estimation type equalizer is an equalizer having the highest ability. However, there is a drawback in a large circuit scale, i.e., it is very calculation-intensive. Therefore, the development of the equalizer is carried forward to reduce a circuit scale without degrading the equalization ability. As one example, there is an equalizer using a delayed decision feedback sequence estimator (DDFSE) in which the maximum likelihood sequence estimator and a decision feedback equalizer (DFE) are combined. Such a delayed decision feedback sequence estimator is described in xe2x80x9cNEC Research and Developmentxe2x80x9d (January, 1997, pp. 74-80).
An example of the delayed decision feedback sequence estimation reception apparatus is described in Japanese Patent Application No. Heisei9-158172 (reference 2: corresponding to Japanese Laid Open Patent Application (JP-A-Heisei 11-8573) opened on Jan. 12, 1999).
FIG. 2 shows a schematic structure of the reference 2. Referring to FIG. 2, when a reception signal is supplied, an impulse response is determined by a transmission path characteristic detector 41 during the reception of a training signal. Also, the amplitudes of the impulse response components are determined by an absolute value calculating unit 42. A summing unit 43 classifies impulse response components into a maximum likelihood sequence estimation region, a decision feedback equalization region and an outside region other than the maximum likelihood sequence estimation region and the decision feedback equalization region. Also, the summing unit 43 determines summations (p, q and r) of the amplitude values for each region. After that, the summing unit 43 calculates the summations p, q and r one after another while shifting each region, to output to a maximum value detector 44. The maximum value detector 44 carries out the calculation of P/(R+xcex1Q) and outputs timings corresponding to the maximum calculation result to a delayed decision feedback sequence estimator 45 (In this example, xcex1=1/7, and the values P, Q and R are the same as defined above).
The delayed decision feedback sequence estimator 45 determines a maximum likelihood sequence estimation region and a decision feedback equalization region of the impulse response components supplied from transmission path characteristic detector 41 in response to the timing signals which are supplied from the maximum value detector 44. The delayed decision feedback sequence estimator 45 carries out a sequence estimation using the impulse response components in those regions, and outputs as the maximum likelihood estimation data.
Next, calculation for determining the optimal region of the impulse response components in the maximum value detector 44 will be described.
All the components of the decision feedback equalization region are ideally canceled through the feedback operation, and do not contribute to improvement or degradation of the estimation ability of the sequence estimator. Therefore, the estimation ability is determined based on the ratio P/R of the components of the maximum likelihood sequence estimation region to the components of the outside region. When the ratio is larger, the estimation ability is higher.
However, the decision feedback equalization region cannot be completely canceled due to errors such as a quantization error so that a component is left as a residual distortion. Therefore, it is possible to say that the estimation ability is higher when the ratio of the components of the maximum likelihood sequence estimation region to the addition of the components of the outside region and the components of the decision feedback equalization region which is weighted by a coefficient xcex1 is larger. That is, the timing when the above-mentioned P/(R+xcex1Q) becomes maximum indicates the optimal sequence estimation region.
In the above Japanese Patent Application No. (Heisei 9-158172), it is described that the calculation to determine such an optimal sequence estimation region is carried out using a simple algorithm in the delayed decision feedback sequence estimator. Thus, the apparatus is made small in size and is reduced in power consumption.
By the way, it is possible to realize an adaptive equalizer with the structure like an adaptive maximum likelihood sequence estimator, if the impulse response components are updated using an adaptive algorithm even in the conventional delayed decision feedback sequence estimator. However, in such an adaptive delayed decision feedback sequence estimator, there is a problem in that the estimation characteristic is deteriorated, depending on the waveform of the impulse response of the transmission path, when the adaptive control is carried out.
FIG. 4 shows an example of an impulse response of the transmission path and the sequence estimation region. Referring to FIG. 4, this example is a 2-wave model in which the maximum likelihood sequence estimation region has 3-symbol length, and the decision feedback equalization region has 3-symbol length. Also, there is a direct wave and a delayed wave with a 4-T (T: symbol period) delay time. There is a case that the level of the direct wave becomes low remarkably due to fading in an actual environment, as shown in FIG. 4. In such a case, the maximum likelihood sequence estimation region is set to contain delayed wave components based on the above-mentioned calculation for determining the sequence estimation region. Thus, the direct wave component is outside the estimation region.
In this way, the delayed decision feedback sequence estimator has the feature in the following point. That is, even when the direct wave and all the delayed wave components fall within the sequence estimation region (6 symbol length in the case), the sequence estimation region is set such that a part of the impulse response components is outside the estimation region.
When a mobile station does not move, the transmission path characteristic changes hardly. In such a situation, a good estimation result can be obtained in the delayed decision feedback sequence estimator, in case where the maximum likelihood sequence estimation is carried out using the delayed wave components with higher levels shown in FIG. 4, rather than a case where the direct wave components with lower levels are used. However, when the adaptive equalization is carried out, the direct wave components out of the estimation region are not reflected to a generated replica signal. Therefore, the direct wave components are contained in the error signal between the replica signal and the reception signal, just as it is. The impulse response components are updated as if the transmission path characteristic changes in spite that the error signal is not generated due to the change of the transmission path characteristic. As a result, the impulse response has no relation with the correct transmission path characteristic so that the estimated characteristic is deteriorated. In other words, in the delayed decision feedback sequence estimator, a region has been set to contain a lot of impulse response components in the maximum likelihood sequence estimation region of the sequence estimation region. Because this is always not the setting to contain a lot of impulse response components in the sequence estimation region, the reliability of the adaptive control replica signal which is generated using the impulse response components in this region become low. As a result, an erroneous adaptive control is carried out and the estimation characteristic has deteriorated.
In conjunction with the above description, a maximum likelihood sequence estimating receiver is described in Japanese laid Open Patent Applications (JP-A-Heisei 5-292138 and JP-A-Heisei 5-292139). In these references, the receiver is composed of a transversal type matched filter 5, a transmission path estimating circuit for estimating impulse responses of a transmission path, and a state estimating circuit 6 for estimating a transmission symbol sequence from the output of the matched filter based on the estimated impulse responses. The transmission path estimating circuit 9 sets a time interval to the estimated impulse response components. The time interval is determined to contain as many sample points as possible in order of the largest amplitude from among the sample points having the maximum amplitudes and to correspond to the number of taps of the matched filter 5. The setting of tap coefficients of the matched filter 5 and the estimation of the transmission symbol sequence in the state estimating circuit 6 are carried out based on only the sample points within the time interval. Thus, impulse response components of the optimum transmission path having the length of NT can be estimated by comparing the signal amplitudes with each other without calculating a summation of squares of signal amplitudes. Alternatively, the time interval is determined to contain as many sample points having amplitudes higher than a predetermined value as possible in order of the largest amplitude from among the sample points having the maximum amplitudes and to correspond to the number of taps of the matched filter 5.
Also, an adaptive equalizer is described in the above Japanese Laid Open Patent Applications (JP-A-Heisei 5-152893 and JP-A-Heisei 5-152894). In these references, the adaptive equalizer carries out an adaptive control based on a maximum likelihood sequence estimating method using a least mean square adaptive algorithm or a recursive least square adaptive algorithm for impulse response estimation of a transmission path. The maximum likelihood sequence estimations are carried out using parameter correction coefficients of the least mean square adaptive algorithms different from each other or forgetting coefficients of the recursive least square adaptive algorithms different from each other for a data or control information symbol interval to one time slot or a plurality of symbols. An estimated value of the transmission symbol sequence is outputted in which the maximum value of pathmetric calculated by a Viterbi algorithm processing section 70 is maximum. Alternatively, the estimation value of the transmission symbol sequence is outputted in which a summation of squares of the outputs of an adder 84 of the transmission path estimating section 80, i.e., a transmission estimation error of the transmission path estimating section 80, for every symbol has a minimum.
Also, an adaptive equalizer is described in Japanese Laid Open Patent Application (JP-A-Heisei 6-216710). In the reference, the adaptive equalizer is composed of an equalizing filter section including delay circuits with taps, a data determining section (12), a tap coefficient updating circuit (23), and a correction signal generating circuit (24) for generating a correction signal based on a reference signal (an output data from the data determining section 12 or a known signal sequence) and a resultant estimation output value determined from the tap coefficients after the update and an input signal of the equalizing filter section used after the update.
Therefore, an object of the present invention is to provide an adaptive equalizer using a delayed decision feedback sequence estimator to have a high equalization ability while the deterioration of estimation characteristic due to an impulse response of a transmission path is suppressed.
In order to achieve an aspect of the present invention, an adaptive equalizer includes an impulse response detecting section, a region specifying section, an adaptive control section and a delayed decision feedback sequence estimating section. The impulse response detecting section detects an impulse response of a transmission path from a training signal. A reception signal includes the training signal and a data signal following the training signal. The region specifying section outputs a region specifying signal used to specify a sequence estimation region and an adaptive control region of the detected impulse response. The adaptive control section updates the detected impulse response using an adaptive algorithm based on the data signal and an estimation result, and outputs the updated impulse response for the determined sequence estimation region. The delayed decision feedback sequence estimating section performs a sequence estimation based on the updated impulse response for the determined sequence estimation region and the data signal to produce the estimation result, and outputs the estimation result to the adaptive control section.
The region specifying section specifies a region where P/(R+xcex1Q) becomes maximum, as the sequence estimation region, and a region where (P+Q)/ R become maximum, as the adaptive control region, where P is a power component of in a maximum likelihood sequence estimation region, Q is a power component of a decision feedback equalization region, R is a power component of a region out of the maximum likelihood sequence estimation region and the decision feedback equalization region, and xcex1 is an optional value.
Also, the adaptive control section includes a replica signal generating section, an error signal generating section and an impulse response updating section. The replica signal generating section generates a replica signal in response to the estimation result and the updated impulse response for the determined adaptive control region. The error signal generating section generates an error signal from the data signal and the replica signal. The impulse response updating section updates the updated impulse response using the adaptive algorithm based on the data signal and the estimation result, and outputs the updated impulse response for the determined sequence estimation region. In this case, the error signal generating section includes a delay section and a subtracting section. The delay section delays the data signal by a predetermined delay quantity to produce a delay signal. The subtracting section generates the error signal corresponding to a difference between the delay signal and the replica signal. Also, the impulse response updating section further sets the delay quantity to the output variable delay section based on a determination delay of delayed decision feedback sequence estimation and a time difference between the sequence estimation region and the adaptive control region.
Also, the replica signal generating section includes a transversal filter which convolutes impulse response for the adaptive control region and the estimation result to generate the replica signal.
In order to achieve another aspect of the present invention, a method of compensating a transmission path distortion, includes:
setting and holding an initial impulse response, a sequence estimation region and an adaptive control region of the detected impulse response based on a training signal, a reception signal including the training signal and a data signal following the training signal;
performing a sequence estimation based on the held impulse response for the held sequence estimation region and the data signal to produce a estimation result;
updating and holding the held impulse response using an adaptive algorithm based on the data signal and the estimation result.
An impulse response of a transmission path is detected from a training signal to set the detected impulse response as the initial impulse response, and the sequence estimation region and the adaptive control region is determined based on the initial impulse response.
In the determining, a region where P/(R+xcex1Q) becomes maximum, as the held sequence estimation region, and a region where (P+Q)/ R become maximum, as the held adaptive control region are determined, where P is a power component of in a maximum likelihood sequence estimation region, Q is a power component of a decision feedback equalization region, R is a power component of a region out of the maximum likelihood sequence estimation region and the decision feedback equalization region, and xcex1 is an optional value.
Also, in the determining, a replica signal is generated in response to the estimation result and the held impulse response for the determined adaptive control region. An error signal is generated from the data signal and the replica signal. The held impulse response is updated using the adaptive algorithm based on the data signal and the estimation result.
In the generating an error signal, the data signal is delayed by a predetermined delay quantity to produce a delay signal. The error signal corresponding to a difference between the delay signal and the replica signal is generated. In this case, in the updating, the delay quantity is predetermined based on a determination delay of delayed decision feedback sequence estimation and a time difference between the sequence estimation region and the adaptive control region.
Also, in the generating a replica signal, the held impulse response for the held adaptive control region and the estimation result are convoluted to generate the replica signal.
In order to achieve still another aspect of the present invention, an adaptive equalizer, includes a delayed decision feedback type sequence estimator, a replica generating unit, an error signal generating unit and an impulse response updating section unit. The delayed decision feedback type sequence estimator performs a sequence estimation based on a held impulse response for a held sequence estimation region and a data signal to produce an estimation result. A reception signal includes a training signal and the data signal following the training signal. The replica generating unit generates a replica signal in response to the estimation result and the held impulse response for a held adaptive control region. The error signal generating unit generates an error signal from the data signal and the replica signal. The impulse response updating unit updates the held impulse response using an adaptive algorithm based on the data signal and the estimation result, and outputs the held impulse response for the held sequence estimation region to the delayed decision feedback type sequence estimator.
The adaptive equalizer may further include an impulse response detector and a region specifying section. The impulse response detector extracts an impulse response of a transmission path from the training signal to set the detected impulse response as the held impulse response in the impulse response updating unit. The region specifying unit determines the sequence estimation region and the initial adaptive control region based on the impulse response to set the sequence estimation region and the adaptive control region in the impulse response updating section.
Also, the region specifying unit determines a region where P/(R+xcex1Q) becomes maximum, as the held sequence estimation region, and a region where (PandQ)/R becomes maximum, as the held adaptive control region, where P is a power component of a maximum likelihood sequence estimation region, Q is a power component of a decision feedback equalization region, R is a power component of a region out of the maximum likelihood sequence estimation region and the decision feedback equalization region, and xcex1 is an optional value.
The error signal generating unit includes a delay unit which delays the data signal by a predetermined delay quantity to produce a delay signal, and an adder generating the error signal corresponding to a difference between the delay signal and the replica signal. In this case, the impulse response updating unit updates the held impulse response using the algorithm based on the error signal and the estimation result.
Also, the impulse response updating sets the predetermining delay quantity based on a determination delay of delayed decision feedback sequence estimation and a time difference between the sequence estimation region and the adaptive control region.
Also, the replica generating unit includes a transversal filter convoluting the held impulse response for the held adaptive control region and the estimation result to generate the replica signal.